Semantic language models with deep neural networks
详细信息 查看全文
- 作者：Ali Orkan Bayer ; aliorkan.bayer@unitn.it" class="auth_mail" title="E-mail the corresponding author ; Giuseppe Riccardi giuseppe.riccardi@unitn.it" class="auth_mail" title="E-mail the corresponding author
- 关键词：Language modeling ; Recurrent neural networks ; Frame semantics ; Semantic language models ; Deep autoencoders
- 刊名：Computer Speech & Language
- 出版年：2016
- 出版时间：November 2016
- 年：2016
- 卷：40
- 期：Complete
- 页码：1-22
- 全文大小：1946 K
文摘
- •
We incorporate semantic features into LMs via theory of frame semantics.
- •
We use deep autoencoders for noise robust encoding of semantic context.
- •
Deep semantic encodings on target words are more noise robust.
- •
Semantic LMs (SELMs) have a better recognition performance on target words.
- •
SELMs have a better understanding performance on the frame identification task.